| From: | Staf Verhaegen |
| Date: | 14 Sep 2000 at 13:50:49 |
| Subject: | Re: AMIOPEN: Re: |
John Niclasen wrote:
>
> Jim Peters wrote:
> >
> > On Thu, Sep 14, 2000 at 08:25:25AM +0200, Juan Carlos Marcos Rodr�guez wrote:
> > > I remember the Dynamo project, a shocking technique for a CPU interpreting
> > > code it can already run, because its native. Major speedups are reported.
> > > The Arstechnica website (a pretty good one, by the way) had a very nice
> > > feature about it (http://arstechnica.com/reviews/1q00/dynamo/dynamo-1.html).
> >
> > This is very very interesting. This suggests that a Dynamo-type
> > translator working at run-time translating the VP code to optimised
> > native code could do a better job than any amount of compile-time
> > optimisation to native code. If a technology like Dynamo comes to the
> > Amiga (or if we already have it !), then VP code will actually be
> > faster than native code in many cases. Certainly something to think
> > about.
>
> I agree, this is very interesting! And throw in an add-on in the
> translator, and you could end up solving the MP (Memory Protection)
> problem, instead of the traditional MMU way.
>
Not if you allow programming languages with pointers (e.g. C). You either have
to check every access through the pointer (or check every pointer when it's
value changes) or use the MMU and I know which is the fastest.
Staf.
+----------------------------------------+-----------------------------+
|Staf Verhaegen (staf.verhaegen@imec.be) |ADRESS: IMEC vzw. - ASP/LITHO|
|tel: 016/ 281 783 | Kapeldreef 75 |
|fax: 016/ 281 214 | 3001 Leuven (Belgium)|
+----------------------------------------+-----------------------------+
For every tool there are at least 2 uses: the one it was designed for
and the other for which it wasn't.
Subscribe/Unsubscribe: open-request@amiga.com
Amiga FAQ: http://www.amiga.com/faq.html